The advent of cutting-edge genomic technologies has dramatically revolutionized modern clinical genetics, enabling comprehensive insights into contemporary disease states with unprecedented resolution. However, these technologies have predominantly been applied to fresh or recently collected genetic material, leaving a vast reservoir of older clinical specimens largely untapped. This limitation curtails our ability to dissect the historical genomic landscape of diseases and understand how genetic factors influencing disease susceptibility have evolved over extended periods.

Harnessing a novel approach inspired by methodologies pioneered in the field of ancient DNA research, scientists at the University of Chicago have adapted sophisticated techniques to robustly recover and sequence degraded DNA from formalin-fixed, paraffin-embedded (FFPE) medical samples dating back nearly a century. This approach bridges the gap between molecular archaeology and modern medical genomics, enabling exploration of genetic changes in tumor biology across decades.

The research was unveiled in a forthcoming presentation at the prestigious Association for Molecular Pathology (AMP) 2025 Annual Meeting & Expo in Boston, a gathering of leading molecular diagnostic experts aimed at pushing the boundaries of disease understanding through genomic science. The team meticulously selected colorectal cancer samples spanning an extraordinary temporal range—from 1932 to 2023—to investigate evolving genetic landscapes and microbial constituents within tumor microenvironments.

Colorectal cancer’s increasing incidence among younger adults, a demographic historically at lower risk, poses a perplexing epidemiological enigma. Contemporary statistics reveal a troubling trend: a 35-year-old today faces twice the likelihood of developing colorectal cancer compared to their counterpart in 1985. A comprehensive temporal genomic analysis promised to shed light on potential genetic drivers underpinning this shift.

Retrieval of actionable genetic material from the FFPE specimens required precise optimization of DNA extraction techniques. The research team innovatively fine-tuned steps for meticulous removal of paraffin wax and chemical preservatives—agents known to complicate nucleic acid recovery. Their tailored protocols maximized yield of usable DNA fragments, a critical and delicate step given the high degree of DNA fragmentation typical of archival tissue samples.

To reconstruct meaningful genomic data from these fragmented strands, the researchers employed a bespoke bioinformatics pipeline rooted in digital tools originally designed to handle ancient DNA. These tools adeptly aligned damaged and incomplete sequences with the human reference genome, overcoming challenges associated with post-mortem DNA damage such as cytosine deamination and strand breaks, allowing for accurate genetic variant calling in an otherwise challenging data landscape.

The pipeline incorporated whole-genome sequencing complemented by a targeted gene-enrichment strategy focusing on a panel of oncogenes and tumor suppressor genes relevant to colorectal cancer (using the OncoPlus panel). Crucially, protocols were modified to preserve ultra-short DNA fragments, which are traditionally discarded during library preparation, ensuring the recovery of maximum genomic information from severely degraded samples.

Beyond human genomic data, the team probed the metagenomic component embedded within the tumor tissues. Analysis of non-human DNA sequences revealed presence not only of commensal gut microbiota but also specific bacterial species previously implicated in colorectal carcinogenesis. This innovative inclusion reveals the multifaceted nature of tumor microenvironments, where microbial constituents may influence oncogenic processes and disease progression.

Dr. Alexander Guzzetta, M.D., Ph.D., leading the initiative alongside ancient DNA expert Maanasa Raghavan, Ph.D., remarked on the groundbreaking nature of this dual genomic and microbial analysis. He emphasized the significance of detecting cancer-associated bacterial species retrospectively across decades, a window that could elucidate shifts in tumor microbiomes and their potential role in the epidemiological transition of colorectal cancer.

The implications of this research extend far beyond colorectal cancer. The demonstrated capacity to recover and decipher genetic information from long-preserved clinical samples paves the way for retrospective analyses of numerous diseases. Guzzetta envisions a future where molecular historians of medicine can unravel the genomic evolution of infectious pathogens, inherited disorders, and cancer types with temporal depth previously unattainable.

Fundamentally, this methodology offers an unprecedented tool for molecular epidemiology and evolutionary pathology, bridging the gap between historic medical archives and cutting-edge genomic science. The ability to track genetic mutation patterns, microbial interactions, and disease susceptibility shifts longitudinally could revolutionize disease prevention, diagnosis, and treatment paradigms.

The team’s work will be highlighted in both a platform presentation and poster session at the AMP 2025 event, scheduled for November 13 at the Thomas M. Menino Convention and Exhibition Center in Boston. Guzzetta will be available to engage with the scientific community and media to discuss the technical intricacies and transformative potential of this research.

As attention turns increasingly to the dynamic interplay between human genetics, microbiomes, and disease, this pioneering approach heralds a new era where historical genetic data fragments, once lost to time, become key to unlocking medical mysteries of the past, present, and future.

Subject of Research:
Advancements in sequencing degraded genetic material from archival medical samples to study disease evolution, with a focus on colorectal cancer genetics and tumor microbiomes.

Article Title:
Decoding Nearly a Century of Cancer Genomics: Ancient DNA Techniques Revitalize Colorectal Tumor Analysis

News Publication Date:
2025 (Ahead of AMP 2025 Annual Meeting & Expo)

Web References:
Association for Molecular Pathology 2025 Annual Meeting & Expo

Keywords:
Pathology, Disease susceptibility, Ancient DNA, DNA damage, DNA, Genetic material, Genetic analysis, Colorectal cancer, Colon cancer

Proteomics, the large-scale study of proteins, involves collecting enormous quantities of protein data that scientists use to compare against samples. These data repositories serve crucial functions, such as enabling clinicians to identify various diseases, evaluate treatment efficacy, and pinpoint pathogens in clinical specimens. However, as researchers at the Technical University of Denmark (DTU) and their collaborators highlight, existing tools still face considerable obstacles. Timothy Patrick Jenkins, an Associate Professor at DTU Bioengineering, notes that databases often lack comprehensive coverage, making it vital for researchers to identify the most relevant resources for their specific inquiries.

Moreover, he points out that deep searches through these databases are not only time-consuming but also computationally demanding, which can impede research progress. The challenge becomes even greater when scientists attempt to identify proteins that are yet to be registered in existing databases, underscoring the necessity for more innovative solutions in the field. Many teams have sought to develop de novo sequencing algorithms to improve accuracy and minimize computational demands, but their accomplishments have been described as “underwhelming” by Jenkins and his partners.

In their recent publication, the team unveils their two pioneering AI models, InstaNovo and InstaNovo+, which are now accessible to researchers through InstaDeep’s website. These tools aim to elevate search precision dramatically and offer a novel approach to proteomic data analysis. According to the research engineer Kevin Michael Eloff, one of the co-first authors, combining these models yields a performance that surpasses the previous state-of-the-art benchmarks. Crucially, these models are versatile enough to address challenges across various research domains involving proteomics, showcasing their broad applicability.

The researchers put their models to the test in several key scenarios, seeking to illustrate their robust performance and unique capabilities. For instance, when applied to analyze wound fluid from patients suffering from venous leg ulcers, a condition notoriously difficult to manage, InstaNovo models were able to identify ten times as many sequences as traditional database searches. This result included the detection of both E. coli and the multidrug-resistant bacterium Pseudomonas aeruginosa, demonstrating the models’ usefulness in clinical contexts.

Furthermore, another critical application of the InstaNovo models involved analysis of small protein fragments known as peptides. These peptides play a pivotal role in aiding the immune system to recognize infections and diseases, including cancer. Utilizing InstaNovo, researchers were able to identify thousands of previously undiscovered peptides that traditional methods failed to catch. This breakthrough is particularly significant as such peptides represent potential targets for immunotherapy, opening doors for more effective personalized cancer treatments.

The implications of InstaNovo extend beyond mere ingenuity in the field of medical science. Konstantinos Kalogeropoulos, co-first author and Assistant Professor at DTU Bioengineering, emphasizes that the models contribute to a deeper understanding of complex biological interactions. They can significantly enhance microbiome identification and improve personalized medicine applications, particularly in cancer immunology. When faced with complex scenarios where unknown proteins or pathogens are present, InstaNovo and its enhancements can catalyze significant advances in how researchers perceive and tackle biomedical challenges.

Additionally, the paper outlines six further cases showcasing how the InstaNovo models can refine therapeutic sequencing, promote novel peptide discovery, recognize unreported organisms, and markedly improve proteomics searches. These potential enhancements transcend the medical arena, as Jenkins conveys. He suggests that adopting these tools illuminates our understanding of the biological world, impacting fields as diverse as plant science, veterinary science, industrial biotechnology, environmental monitoring, and even archaeology. With these innovations, researchers are expected to gain critically needed insights into protein landscapes that were previously unreachable.

A key distinguishing feature of InstaNovo is that it operates as a transformer-based model specifically formulated for de novo peptide sequencing. This model’s ability to translate mass spectrometry data into peptide sequences with exceptional precision addresses gaps that traditional methods left unfilled. Unlike conventional techniques reliant on pre-existing databases, InstaNovo is designed to recognize and document novel peptides, enriching the proteomic discovery pipeline.

Complementing InstaNovo is InstaNovo+, a diffusion-based iterative refinement model that enhances sequence accuracy by emulating the meticulous refinement process researchers typically undertake manually. Beginning with an initial peptide sequence—either derived from InstaNovo or generated randomly—InstaNovo+ methodically improves the predictions, effectively refining the accuracy and minimizing false discovery rates. This dual approach combines precise predictions with extensive exploration, positioning InstaNovo and InstaNovo+ as a revolutionary pair in peptide sequencing endeavors.

The innovative methodologies encapsulated within InstaNovo and InstaNovo+ reflect a significant leap forward in the proteomics landscape. With their capabilities to improve both the breadth and depth of protein analysis, these models are set to accelerate biological discoveries at an unprecedented scale. As researchers from DTU, Delft University, and InstaDeep navigate the ever-evolving landscape of protein science, their contributions will no doubt yield impactful shifts in not only healthcare but also diverse industries utilizing proteomic data.

The collaboration between cutting-edge AI technology and biotechnological research highlights a transformative era where computational prowess meets biological inquiry, resulting in far-reaching implications for human health and understanding of life itself. Central to this evolution are the breakthroughs enabled by InstaNovo and InstaNovo+, which provide both researchers and practitioners a contemporary toolkit to unravel the complexities of protein functions and interactions, ultimately redefining the boundaries of what is possible within protein science.

Both InstaNovo and InstaNovo+ stand as integral resources for academic and commercial researchers seeking to enhance their understanding and application of proteomics. The ability to identify previously uncharacterized peptides and organisms places these models at the vanguard of scientific discovery, offering a clearer lens through which the intricacies of biological systems can be explored. As we anticipate the transformative potential of these AI advancements, it becomes clear that the marriage of machine learning and protein science promises to unlock new dimensions of insight into the molecular structures that govern life.

In summary, through continuous refinement and innovative approaches, the advancements represented by InstaNovo and InstaNovo+ herald a new chapter in the quest for knowledge in protein science. As applications extend across various fields, these AI-driven models are poised to equip researchers with the tools necessary to unravel the mysteries of proteins, pathogens, and various biological phenomena.

Subject of Research: The development and application of AI models, InstaNovo and InstaNovo+, in protein science.

Article Title: InstaNovo enables diffusion-powered de novo peptide sequencing in large scale proteomics experiments.

News Publication Date: 31-Mar-2025

Web References: https://www.instadeep.com/2025/03/enhancing-peptide-sequencing-with-ai/

References: DOI: 10.1038/s42256-025-01019-5

Image Credits: Source from InstaDeep.

Keywords: AI, InstaNovo, proteomics, peptide sequencing, drug discovery, personalized medicine, machine learning, biotechnology, protein science, E. coli, Pseudomonas aeruginosa, personalized cancer treatments.